• No results found

Cover Page The handle http://hdl.handle.net/1887/73814 holds various files of this Leiden University dissertation. Author: Joswig, A. Title: The Majang Language Issue Date: 2019-06-04

N/A
N/A
Protected

Academic year: 2021

Share "Cover Page The handle http://hdl.handle.net/1887/73814 holds various files of this Leiden University dissertation. Author: Joswig, A. Title: The Majang Language Issue Date: 2019-06-04"

Copied!
53
0
0

Bezig met laden.... (Bekijk nu de volledige tekst)

Hele tekst

(1)

Cover Page

The handle

http://hdl.handle.net/1887/73814

holds various files of this Leiden University

dissertation.

Author: Joswig, A.

(2)

51

Part II: Phonological Inventory and

Orthography

This section introduces the inventories of the Majang phonology – vowels, consonants and suprasegmentals. It also displays the rules which govern the use of the various elements of these inventories.

In this section, data presented without any kind of bracketing is understood to be in its surface-phonemic representation, transcribed according to the phoneme inventories shown in tables 1 and 3. Only when further phonetic detail needs to be shown for greater clarity, phonetic data is supplemented in square brackets […]. On the other hand, in some places information is given about the individual morphemes making up a word. These are always shown in their underlying representation, and the brackets {…} are used to sur-round such lexical units. This is different from the other parts of this lan-guage description, where the underlying representation is shown without brackets.

Individual sounds are shown by /x/ to be phonemic, by [x] to be phonetic and by <x> to be in orthographic representation.

II.1 Vowels

This section first assesses the vowel inventory, then shows proof of contrast for the various identified phonemes, and finally looks into distributional re-strictions.

II.1.1 Phonemic inventory

(3)

Front Central Back

High i

High-Mid e o

Low-Mid ɛ ɔ̀

Low à

Table 1: vowel phonemes

Bender differs from Unseth and Moges on the Majang vowel inventory. He (1983, p. 114) felt only confident to present six vowels; he noticed the seventh vowel /ɔ̀/, but did not see enough contrastive evidence to confirm its phonemic status. Unseth (1991, 2007) proposed nine vowels, and Moges (2008) even ten vowels for Majang. The chart presented above is the same as the one proposed in Joswig (2012) and Getachew (2014).

In a 2879-word text sample, // is the most frequent vowel with 1336 occur-rences, closely followed by /ɛ/ with 1237 occurrences. /i/ appears 867 times, /o/ 506 times, and /ɔ̀/ 401 times. The least-frequent vowels are /e/ with 347 occurrences and // with only 262 occurrences. This means that the vowel // occurs five times more often than the vowel //.

II.1.2 Contrasts and phonetic realizations

Bender (1983, p. 114) was not able to prove contrast between /o/ and /ɔ̀/. This was merely due to lack of data, as such contrast does exist in the lan-guage: moːrɛŋ‘he boils’ vs. mɔ̀ːrɛŋ ‘it is shriveled’ is an incontestable mini-mal pair between the two vowels. Such a contrast would also be expected for reasons of symmetry.

Example II.1: vowel contrasts6

i vs. à tí*mŋ he wounds tíjɛŋ he hears

tà*mŋ it drips tàjɛŋ he opens

i vs. ɛ kùrí tree, sp. pàːríŋ he tries

kùrɛ hunting net pàrɛŋ he chops

i vs. e kondí fish trap tíːmà n cloud

konde bottle temà ːn firewood

6 In this and all other following contrast charts, the words are presented in their citation form.

(4)

Vowels 53 i vs. ɔ̀ íreL footprint cɔ̀ːbí hoof

ɔ̀rɛːL ancestor ɔ̀bɔ̀ knot

i vs. o ílànL udder tíjɛŋ he hears

oːlàn ability toːjɛŋ he pierces

i vs. ù kíríL thread tíːɟeŋ he takes revenge

kùrí tree, sp. tùːɟeŋ he roasts

e vs. à emɗL canoe jímeL cemetery

àmɗL abdomen jímàL back

e vs. ɛ ɟeːmɛL before ɗejŋ cooking stone

ɟɛːmɛL tree, sp. ɗɛjíŋ he desires

e vs. ɔ̀ remɛL duty keːr courtyard

rɔ̀mɛL proverb kɔ̀ːr middle

e vs. o ceːd here keːr courtyard

coːd there koːr ditch

e vs. ù teːl lake ɗeːk under

tùːl five ɗùkL forest

ɛ vs. à mɛlɛŋ he arrives ɓɛ*rŋ he crushes

màlɛŋ he strikes ɓà*rŋ he forbids

ɛ vs. ɔ̀ wɛːr storm ɗepɛL lion

wɔ̀ːr feast ɗepɔ̀ entrance hall

ɛ vs. o ɛltL grassland kɛ*ŋ he pounds

oltL fish ko*ŋ he gathers

ɛ vs. ù korɛŋ he peels kùrɛ hunting net

korùŋ he closes kùrù foam

à vs. ɔ̀ bàdɛŋ it disappears pàːkɛŋ it is hot

bɔ̀dɛŋ he escapes pɔ̀ːkɛŋ he tears

à vs. o tàːjɛŋ he harvests kàwɛŋ he bites

toːjɛŋ he pierces kowɛŋ it is sour

à vs. ù màɲàL sister tà ːr meat

mùɲà earthworm tùr garbage dump

ɔ̀ vs. o kɔ̀ːr middle pɔ̀ːcɛŋ he praises

(5)

ɔ̀ vs. ù bɔ̀lɔ̀ŋùrkùŋ he is old ɗepɔ̀ entrance hall

bùlùnkùr bubble tàɗàpùL ashes

o vs. ù goːmoj trap toːr smoke

goːmùjL tree, sp. tùr garbage dump

There is significant phonetic variation for most of the vowels. The individual vowels show an enormous bandwidth with respect to their formants, which is caused by various phonological factors, such as syllable structure, position in the word, and consonants in the syllable. Therefore it is easily possible to hear all kinds of vowels in the raw phonetic data. This obviously prompted Unseth and Moges to claim the row of [-ATR] high vowels /ɪ, ʊ/ for Majang. Current speakers of Majang, however, do not perceive any difference be-tween [ɪ] and [i], or between [ʊ] and [], respectively. These phones repre-sent the same sound unit to them. Words like díldílŋ ‘he is fat’ or gírgíɗǐːŋ ‘he rolls’ certainly have instances of [ɪ] in them (the latter only in the first syllable). They can be explained as the positional variant of /i/, when it hap-pens to be short and preceding a lateral or flap consonant. More con-spicuous, however, are cases where the unsuspecting ear perceives an [e] – these could be analyzed as instances of /ɪ/. The two sounds are difficult to distinguish in many languages with nine- or ten-vowel ATR-contrast systems (Casali, 2008, p. 509). I have indeed heard words like péːt ‘girl’ or kéːr ‘courtyard’ with different vowel qualities from different speakers. Some of these pronunciations prompted me to transcribe them as [pɪːt] and [kɪːr]. When pressed hard for these differences, the speakers did not hear them. They agreed that the difference, if there was any, is of no consequence at all to the meaning of the word, and all variations are perfectly acceptable. The same is true for words like ɲón ‘lie’ or ɗóː ‘land’, where there is a similar potential confusion between [o] and [ʊ]. Some phonetic variation can be heard, but it does not matter at all to the speakers.

(6)

Vowels 55 The tenth vowel of Moges’ description, which he transcribed as [ʌ], does not exist in Majang as an independent sound unit either. There is, however, an enormous bandwidth of //-like vowels phonetically. The first formant (F1) can be as low as 600 Hz and as high as 900 Hz with the same speaker, and the second formant (F2) can be between 1250 and 1650 Hz. Again, this con-siderable variation can be explained by environmental factors. A sonorant consonant following the vowel in the same syllable has a lowering effect on the first formant of the phoneme. A long /àː/ has a higher chance of being pronounced with a high F1. Palatal glides or nasals in the same syllable have a tendency of raising the second formant. Short /ɛ/ is often pronounced as [ə], that is as a central unrounded mid vowel.

To illustrate the seven vowels of Majang, Figure 3 below (taken from Joswig 2012, p. 273) gives an indication of the first two formants of each vowel. To make things comparable, the chart represents only the vowels of monosyl-labic nouns and verbs (3rd person singular) of the language (a total of 96 vowel tokens). Each symbol in figure 3 stands for one token. The vertical di-mension displays the frequency of F1 in Hertz, and the horizontal didi-mension shows the frequency of F2.

(7)

In such a controlled environment, the seven vowels show up as surprisingly discrete groupings. Except for the vowel //, there is no significant variance with respect to F1. Casali (2008, p. 507f) states that “[…] all else equal, a [+ATR] vowel will have lower F1 than its [-ATR] counterpart. […] F1 has consistently been found to be a very robust cue for distinguishing [+ATR] vowels from their [-ATR] counterparts […].” Therefore, it would be very difficult to interpret a nine- or ten-vowel system from the picture presented here. If there are only seven vowel phonemes in monosyllabic words, there is no reason to expect a larger number of vowel phonemes in polysyllabic words, even if the vowels in these polysyllabic words show a more signifi-cant phonetic variation regarding F1 and F2.

II.1.3 Vowel length

Vowel length plays an important role in Majang. This is typologically quite different from the Southeast-Surmic languages, which seem to make use of vowel length only sparingly; in both Mursi and Suri, the few long vowels are attributed to the loss of intervocalic consonants (Bryant, 2013, p. 28f; Mütze, 2014, p. 39f). The distinction between long and short vowels cannot be ana-lyzed in these terms in Majang. It is frequently used and lexically contrastive, as can be easily confirmed by numerous minimal pairs:

Example II.2: short and long vowels

a)ùteŋ {ùt-ɛ=ŋ} he drinks vs. ùːtéŋ {ùːt-ɛ=ŋ} it rusts

b)olàn {olàn} husband vs. óːlánL {óːl-ànL} be able (INF)

c)gɔ̀gɔ̀j {gɔ̀gɔ̀j} he crawls vs. gɔ̀ːgɔ̀j {gɔ̀ːgɔ̀j} ford

d)tàjàːŋ {tàj-à=ŋ} I open vs. tàːjáːŋ {tàːj-á=ŋ} I harvest e)kɛtɛŋ {kɛt-ɛ=ŋ} he cuts vs. kɛːtɛŋ {kɛːt-ɛ=ŋ} he scatters f)menɛŋ {men-ɛ=ŋ} he twists vs. meːnɛŋ {meːn-ɛ=ŋ} he steers

g)ín {ín} HORT vs. íːn {íːn} you

(8)

Vowels 57

Figure 4: duration of short vowel in ùtéŋ ‘he drinks’

Figure 5: duration of long vowel in ùːtéŋ ‘it rusts’

(9)

speakers would readily agree that the first vowel in ‘drink’ is short, and in ‘rust’ it is long, there are some suffixes where the evidence is less clear. For example, regarding the 1S suffixes - and -à, my consultants sometimes

pro-nounced them long, sometimes short. In the same way other suffixes, usually those ending in a vowel, had conflicting length evaluations. A good example of that is the impersonal suffix -ɛː, for which the consultants insisted on a long vowel for most words. In some cases, however, they would always tell me to transcribe it as a short vowel, as in ríjɛL ‘he is called’. In other cases where the consultants insist on a short vowel, it can clearly be seen that the vowel goes back to a long vowel underlyingly. This is the case with the short version of the TF.3S suffix -ge ːɗ, which is -geL according to the perception of

the speakers. Similar evidence can be found for the infinitive markers -ɛːL and -íːL, which again are truncated versions of a sequence of long vowel

plus /ɗ/ (see section IV.2.2.1 for more details). Indeed the phonetic evidence is inconclusive, as the duration of these vowels is somewhere in the span be-tween 80 and 110 ms, as in the following illustration of the word tònúrgéL ‘he shouted at them’, taken from a recorded text.

Figure 6: duration of -geL in tònúrgéL

Accordingly, there is some lack of clarity regarding the length of vowels in some morphemes, and some of these could probably have been represented differently in this study.

(10)

Vowels 59 Example II.3: aspectual categories expressed by vowel length

toroꜜkàːŋ I step vs. toroːꜜkáːŋ I trample

Some verbs display stems with changing vowel lengths across the same paradigm:

Example II.4: stems with different vowel lengths in verb paradigms gà*jŋ he imitates vs. gàːjàːŋ I imitate

The same is true for nouns, where length is used to express number and case differences:

Example II.5: stems with different vowel lengths in noun paradigms a) kàːrí coffee leaf (SG.ABS) vs. kàríkL coffee leaves (PL.ABS)

b) tɔ̀ ːn child(SG.ABS) vs. tɔ̀ n child(SG.ERG)

Short vowels are considerably more frequent than their long counterparts. In the 2879-word text sample, 3817 short vowels are balanced by 1139 long vowels, a ratio of more than 3:1. But this ratio is not evenly spread out for all vowels. Whereas 243 short vowels /e/ occur compared to 104 long counter-parts, a ratio of 2.33:1, against 245 short vowels // there are only 17 long vowels /ùː/, a ratio of 14.4:1. The other five vowels are much closer to the established average. One possible explanation for the surprisingly high ratio between the // vowels is that many of the short ones go back to a vowel /i/ undergoing labial harmony, which does not affect long vowels //.

There is a difference between a long vowel and two identical short vowels meeting at morpheme boundaries. All previous examples show long vowels belonging to one morpheme. They are then pronounced as just that – long vowels. But there are other possibilities:

Example II.6: apparent long vowel resulting from two adjacent morphemes ɛrcɛɛL milk < {ɛrcɛ -ɛL}

milk -PL.ABS

(11)

how-ever, the two-vowel sequence is often indistinguishable from a long vowel. So example II.6 may be pronounced as either [ɛrsɛː], [ɛrsɛjɛ] or [ɛrsɛʔɛ], with no difference in meaning. Throughout this work, I show both vowels spelled out when they belong to two different morphemes, and use the phonetic length symbol /ː/ to indicate a long vowel belonging to only one morpheme. The Majang language allows vowel sequences involving long vowels. Example II.7: vowel sequences

a) {dàràː-ɛːL}

despise-IMPS

it is despised

b) {àːbɛ-ɛːk-ɛ}

fig_tree-PL-LOC

at the fig trees

Example a) has a long vowel (the impersonal marker) following a long stem vowel. In example b) the long vowel of the plural morpheme follows a short stem vowel of the same quality. If no glottal stop is used to separate the two vowels, the resulting vowel length of /ɛɛː/ appears to not be significantly longer than a regular long vowel /ɛː/, or the union of two short vowels /ɛ/ as in {àːbɛ-ɛL} ‘fig trees’ (PL.ABS).

II.1.4 Distribution

Most of the vowel phonemes can appear anywhere in a word. Vowels can be found word-initially, word-finally and inside a word.

Monosyllabic words with only a short vowel and no coda are very rare. Only function words like ‘with’, ɓáL ‘REMPST’ or the conjunction kɛ belong to that

category. Nearly as infrequent are monosyllabic words ending in a long vowel: ɗoː ‘world’, pɛː ‘soup’ and wàː ‘house (DAT)’ are the only nouns found

in that category. The only verbs encountered with a stem resulting in such words are ŋàː ‘he stinks’, meː ‘it hurts’, dɛː ‘it is red’, kɛː ‘he goes’ and wɛː‘he breathes’.

(12)

Vowels 61 has an effect on the stem vowel (Casali, 2008, p. 500). Neither case can be observed in Majang. On the contrary, the verbal system allows combinations of vowels which would be deemed impossible in a language with ATR-based vowel harmony. Here are some examples:

Example II.8: vowel combinations with different ATR values a) ɛːkí ːŋ b) mɔ̀ːꜜrún

{ɛːk-í,ːL=ŋ} {mɔ̀ːr-ín}

draw-1P.DJ=SFT wrinkled-2S.DJ

we draw water you are wrinkled

Both the stems ɛːk and mɔ̀ːr are completely stable, regardless of the following suffixes. The same is true about the 1P.DJ-suffix -í,ːL, which undergoes some

tonal variation, but remains stable with respect to the vowel. The suffix for the 2nd person singular disjoint is -ín, as in làːrín ‘you lose’, but when

follow-ing a labial consonant or a syllable with a round (or labial) vowel, it is rounded to -ún, as in the previous example. No vowel changes can be attributed to an ATR vowel harmony.

Even within the roots of nouns or verbs, it is possible to find vowel combi-nations which defy any attempt to describe them in terms of an ATR-based vowel-co-occurrence restriction. Roots like eːbɔ̀d ‘heal’ and éːmɛj ‘honor’ should not be encountered in a language with ATR-based vowel harmony.

V1/V2 i e ɛ ɔ̀ o ù total

i 64 24 36 48 9 14 1 196 e 28 40 15 32 1 4 2 122 ɛ 43 4 72 33 4 6 2 164 à 90 19 81 119 3 15 26 353 ɔ̀ 21 17 26 16 6 86 o 24 12 25 36 26 27 150 ù 10 14 13 8 1 7 24 77 total 280 113 259 302 34 72 88 1148 Table 2: vowel co-occurrences in polysyllabic words

(13)

column 1 is followed by a vowel of row 1, regardless of whether this vowel is part of the stem or of a suffix.

Some conclusions can be drawn from this table: almost any vowel can be followed by any other vowel, except for the two back-mid vowels /o/ and /ɔ̀/, which were never found following each other, and the sequence /ɔ̀-e/. These look like structural constraints which may indeed be going back to an old ATR-based vowel-harmony system at an earlier stage of the language’s his-tory. For many vowels, the preferred combination appears to be with the same vowel. The vowels /à, ɛ, e, i, ù/ show a high occurrence as follow-up vowels in a vowel sequence, which is best explained by their prominent use in suffixes. The vowel /ɔ̀/, although not at all infrequent in the language, has a disproportionately low occurrence rate as a second vowel, because it is rarely used in suffixes, except in demonstratives – and each different demon-strative is only counted once in the above chart.

Some combinations are very infrequent. The one token filling the slot /e-ɔ̀/ is the short relative pronoun écɔ̀. The slot /i-ù/ is tàken by the complex verb pàːrírkúnɗɔ̀L ‘while you

PL are trying’. This same word is also the only place

where the sequence /ù-ɔ̀/ was encountered, as this contains the rare 2P marker

-ɔ̀L.

The picture is not very different when only monomorphemic stems are taken into account. The vowel /ɛ/ can still follow /e/ inside a root, but not the other way around. The only other restrictions in addition to those in Table 2 con-cern the vowel //, which never seems to follow any of the vowels /i, ɛ, ɔ̀/ inside of roots.

Therefore, although there are some apparent restrictions regarding the co-oc-currence of vowels, there is enough evidence discounting any ongoing ATR-based vowel harmony in Majang. The picture is, of course, complicated by the fact that four Proto-Surmic vowels /ʊ, o, ɪ, e/ (Moges, 2002, p. 209) have collapsed into the two vowels /e/ and /o/, so that what now may look like a violation of ATR-based vowel harmony could in fact go back to a difference that was present in an earlier stage of the language. For example, a root such as eːbɔ̀d ‘heal’ is likely to go back to a proto-language root *ɪːbɔ̀d, which would not violate an ATR-based vowel-harmony system7. Still, it is

surpris-ingly difficult to find traces of the old ATR-based vowel harmony in

Ma-7 This word is probably an old causative of bɔ̀d ‘be well’. See section V.5.3 for more on the

(14)

Vowels 63 jang8. A good example for this are the two vowels /e/ and /ɛ/. They look very

much like [+ATR] and [-ATR] counterparts of the same vowel, so one should expect them to show some kind of distribution which follows old ATR-based patterns. Indeed, the two vowels alternate in some suffixes which contain them, but based on a principle unrelated to ATR: /e/ only oc-curs in words with a high vowel in the previous syllable, whereas /ɛ/ shows up in all other environments. Therefore the variation can be described in terms of a height assimilation. This becomes clear from the following ex-amples:

Example II.9: height assimilation

a) cùːwe*rŋ b) cùːweŋ

{cùːw-er=ŋ} {cùːw-e=ŋ}

sting-3P.DJ=SFT sting-3S.DJ=SFT

they sting it stings

c) ɔ̀ːjɛ*rŋ d) ɔ̀ːjɛŋ

{ɔ̀ːj-ɛr=ŋ} {ɔ̀ːj-ɛ=ŋ}

obstruct-3P.DJ=SFT obstruct-3S.DJ=SFT

they obstruct he obstructs

e) rùːme*rŋ f) rùːmeŋ

{rùːm-er=ŋ} {rùːm-e=ŋ}

decide-3P.DJ=SFT decide-3S.DJ=SFT

they decide he decides

g) oːdɛ*rŋ h) oːdɛŋ

{oːd-ɛr=ŋ} {oːd-ɛ=ŋ}

difficult-3P.DJ=SFT difficult-3S.DJ=SFT

they are difficult it is difficult

i) tíːɟe*rŋ j) tíːɟeŋ

{tíːɟ-er=ŋ} {tíːɟ-e=ŋ}

avenge-3P.DJ=SFT avenge-3S.DJ=SFT

they avenge he avenges

k) ɓeːcɛ rŋ l) ɓeːcɛŋ

{ɓeːc-ɛr=ŋ} {ɓeːc-ɛ=ŋ}

touch-3P.DJ=SFT touch-3S.DJ=SFT

they touch he touches

8 Examples IV.107 and IV.119 of two infinitive allomorphs may show traces of the old

(15)

m) àːmɛ*rŋ n) àːmɛŋ

{àːm-ɛr=ŋ} {àːm-ɛ=ŋ}

yawn-3P.DJ=SFT yawn-3S.DJ=SFT

they yawn he yawns

As can be seen, the two allomorphs of the morphemes for 3rd person singular

and 3rd person plural in this class of verbs obviously depend on the preceding

stem vowel. The /ɛ/-variant can co-occur with any vowel, as long as it is not high. High vowels require the presence of the /e/-variant of the respective suffix (see section II.5.4). Again, words like oːdɛŋ or ɓeːcɛ rŋ would not be expected in an ATR-based vowel-harmony system.

The behavior of the 2S suffix in example II.8b) lends itself to a description in

terms of a rounding harmony: Example II.10: rounding harmony

a) íbáːlí rŋ b) ɓokotù*rŋ {íbáːl-ir=ŋ} {ɓokot-ùr=ŋ}

play-3P.DJ=SFT kill-3P.DJ=SFT

they play they kill

For a detailed description of the rounding harmony process, see section II.5.3.

II.2 Consonants

Just like the previous section on vowels, this section on consonants first as-sesses the consonant inventory, then shows proof of contrast for the various identified phonemes, and finally looks into distributional restrictions.

II.2.1 Phonemic inventory

(16)

Consonants 65 Mütze (2014, p. 26) counts 21 consonants for Mursi, and Bryant (1999, p. 16) 22 for Suri, which is only slightly more than found in Majang. Majang uses four places of articulation, and a maximum of six different manners of articulation can be found for the alveolar sounds. Only the obstruents and the nasals cover all four places of articulation.

Labial Alveolar Palatal Velar

Voiceless Obstruents p t c k Voiced Obstruents b d ɟ g Implosives ɓ ɗ Nasals m n ɲ ŋ Oral Sonorants w r j Lateral Sonorant l

Table 3: Majang consonant phonemes

The consonant inventory of Majang appears to be very tidy at first glance. For each of the four places of articulation, there is a voiced and a voiceless obstruent, as well as a nasal. At the palatal place of articulation, though, this tidiness is achieved only through a notable abstraction from the phonetic facts, which will be discussed and defended in detail in section II.2.2.3 on palatal consonants below.

Bender (1983, p. 116) shows the exact same consonant inventory, but with the addition of the glottal stop [ʔ], which he claims to be a phoneme of the language. Getachew (2014, p. 49) appears to agree with Bender, calling the glottal stop a phoneme; then again, he also seems to be aware of its complete predictability. Unseth (1988a), who otherwise fully shares Bender’s inven-tory, rejects the phonemic status of the glottal stop on the grounds that it only occurs in word-initial position or between a few prefixes9 ending in a

vowel and the following stems. In the analysis proposed here, the glottal stop is only a phonetic device for separating syllables with a vowel onset, and by no means the only device (see example II.6 for a case where the glottal stop may be inserted). The syllabification rules of Majang only allow the inser-tion of a glottal stop when the preceding syllable ends in a vowel, or if the syllable is at the beginning of the word. Therefore the sound is excluded from the phoneme inventory. Apart from this small difference and a few dif-ferent choices for the phoneme characters, the consonant inventory presented

(17)

above is the same as that proposed not only by Bender (1983), but also by Unseth (2007, p. 628). In addition, Getachew (2014, p. 49) includes /h/ in the phonemic inventory of Majang, qualifying this with the observation that it only seems to appear in loan words.

The distinction between fricatives, affricates and plosives is not relevant in the Majang phonological system, as there is no opposition between these at any place of articulation. Implosives are firmly established in the system, al-though only at the labial and alveolar places of articulation. The consonant inventory is completed by the two liquids /l/ and /r/ and by the two glides /j/ and /w/.

The following chart provides a frequency count of all consonant phonemes of Majang based on a 2879-word text sample:

Example II.11: frequency of consonant phonemes rank phoneme count rank phoneme count

1 k 1190 10 l 208 2 n 987 11 ɓ 187 3 r 487 12 j 169 4 c 448 13 ɟ 151 5 t 445 14 d 140 6 g 351 15 w 119 7 ɗ 308 16 b 109 8 m 306 17 p 87 9 ŋ 254 18 ɲ 35

(18)

Consonants 67

II.2.2 Contrasts and phonetic realizations

This section provides proof of contrast between all possible pairs of pho-nemes that can be reasonably suspected of being variants of the same phoneme. Furthermore, information is given on how the phonemes are pro-nounced; if this is different depending on context or environment, informa-tion about allophones and free variainforma-tion is also provided.

II.2.2.1 Labials /p, b, ɓ, m, w/

Most phonemes of the labial place of articulation are pronounced according to the IPA value of their symbols chosen here. The voiceless plosive /p/ is pronounced with a considerable amount of aspiration. The obstruent /p/ may also be realized by some speakers as a labiodental fricative [f] (Bender, 1983, p. 116). The implosive /ɓ/ is just that phonetically: [ɓ]10. In the Godare

variety of Majang, the voiceless plosive /p/ often is not fully released at the end of a word, to the point that it is barely audible at all.

Example II.12: contrasts between labials

p vs. b pedtàn end àpɛːL grandfather

beːdtàn inhabitant àːbɛ fig tree

p vs. ɓ pàkàtíŋ he decreases tɛpɛr shoulder blade

ɓàkàtíŋ he unwraps tɛɓɛrL thunder

p vs. m pàcɛŋ he carves pàɲàL cousin

màcɛŋ he borrows màɲàL sister

p vs. w pàrɛŋ he chops àpɛːL grandfather

wàrɛŋ he looks for àwɛ iron

b vs. ɓ beɗíŋ he sits bàːlɗíːɗíŋ he throws

ɓeɗíŋ he is awake ɓàlɗíɗíŋ he sells

b vs. m àbí ː cloth boːro gecko

àːmí hair moːr anvil

b vs. w àːbɛ fig tree bɔ̀dɛL oil palm

àwɛ iron wɔ̀ːr feast

10See Moges (2006, p. 823ff) for a very careful description of the implosive sounds of Ma

(19)

ɓ vs. m gàɓɛŋ he accuses ɓùːkeŋ he uncovers

gàmɛŋ it is a fetus mùkeŋ he stabs

ɓ vs. w cùɓoj clay ɓeɗíŋ he awakes

cùwo*j eel wenɛ antenna

m vs. w màcɛŋ he borrows kemùŋ straighten

wàːcɛŋ he speaks keːwùŋ sharpen

II.2.2.2 Alveolars /t, d, ɗ, n, l, r/

The voiceless plosive /t/ is realized with noticeable aspiration []. The pho-neme /r/ is pronounced as a flap [ɾ]. The implosive /ɗ/ is pronounced accord-ing to the IPA value of its phonemic character [ɗ]11, and likewise the other

phonemes /d, n, l/ as [d, n, l]. In the Godare variety of Majang, the voiceless plosive /t/ often is not fully released at the end of a word, to the point that it is barely audible at all.

Example II.13: contrasts between alveolars

t vs. d tàːmɛL face tà*mŋ it drips

dàːmɛ yellow dǎːmŋ he chooses

t vs. ɗ tà*mŋ it drips toɲɛŋ he stops up

ɗà*mŋ he eats ɗoɲɛŋ he smears

t vs. n mermetL red pepper àpàtíL breast

mermen notice àpàní current

t vs. l tɔ̀kɔ̀j lazy kàtàmɛ town

lɔ̀kɔ̀j plate pàlàmɛL argument

t vs. r ɗoːtùŋ he harvests mot blind person

ɗoːrùŋ it crows moːr anvil

d vs. ɗ dǎːmŋ he chooses pedíŋ it is consumed

ɗà*mŋ he eats ɓeɗíŋ he is awake

d vs. n doːmɛŋ he allows ŋàdíŋ he is sad

nomɛŋ he follows ŋàníŋ he turns around

d vs. l díːleŋ he carries dɔ̀ːkɛŋ he wanders

líːleŋ he sinks lɔ̀kɛŋ he overturns

(20)

Consonants 69 d vs. r àːdíŋ he washes doːmɛŋ he allows

àríŋ he weaves romíL morning

ɗ vs. n ɓeɗíŋ he is awake wo ːɗ who?

ɓeːníŋ he sews won which?

ɗ vs. l ɗùkL forest oːɗɛŋ it is difficult

lùkL bastard oːlɛŋ he is able

ɗ vs. r ɗàːwùŋ it is bright píːɗeŋ he shivers

ràːwùŋ he singes píreŋ he flies

n vs. l nomɛŋ he follows tɔ̀ ːn new

lotɛŋ he adds bɔ̀ ːl weak

n vs. r tɛpɛnL forehead woníL tree, sp.

tɛpɛr shoulder blade worí money

l vs. r weːlɛŋ he robs tùl rain

weːrɛŋ he alters tùr garbage dump

II.2.2.3 Palatals /c, ɟ, ɲ, j/

The two palatal obstruents /c, ɟ/ show the most phonetic variation of all con-sonants in Majang: the voiceless obstruent /c/ is most often realized as a voiceless alveolar fricative [s], whereas its voiced counterpart /ɟ/ (/j/ in Bender’s inventory) is usually pronounced as a voiced post-alveolar affricate [d͡ʒ]. The voiceless palatal /c/ may be pronounced as a voiceless post-alveo-lar fricative [ʃ] by some speakers, which supports its analysis as a palatal sound in the system. A regular allophonic rule applies for /c/ following the palatal nasal /ɲ/, which turns /c/ into [t͡ʃ]: ɓeɲce [ɓeɲt͡ʃe] ‘today’. The voiced palatal obstruent also has a less frequent variant [ʒ] (voiced post-alveolar fricative).

(21)

would be possible to treat /c/ as an alveolar fricative /s/, as was done by Getachew (2014, p. 48), but this phonetically driven classification of sounds would add another dimension of complexity into the consonant inventory by claiming an opposition between plosives and fricatives, for which there is no further evidence in the language. At the same time, it leaves a glaring gap for the voiceless counterpart of the phoneme /ɟ/, which in all its phonetic realiza-tions is clearly post-alveolar. In effect, the decision to follow Bender’s treat-ment of /c/ as a palatal phoneme implies that the palatal space in Majang is wider than in other languages, including the edge of the alveoles at which the front-most sibilants can be formed. This is quite in line with Trubetzkoy (1939), who claims (p. 35) that the phoneme is best defined as the total of the phonologically relevant features of a sound entity, and that its content en-tirely depends on its place in the system, defined by the distinctive opposi-tions found in the language (p. 39). The same idea is built upon more recent-ly by Clements (2003) in his concept of Feature Economy.

There is, however, no phonological rule in Majang that changes /ɟ/ to /c/ or vice versa, as such a rule would clearly show that both phonemes belong to the same place of articulation. In the same way, there is also no rule that shows /c/ to be associated with the alveolar sounds of the language.

The palatal nasal /ɲ/ and the palatal glide /j/, however, are pronounced, like their IPA symbols suggest, as [ɲ] and [j]12. In the other places of articulation,

the Godare dialect has an unreleased variety of the voiceless obstruents. This is not applicable for the palatal obstruent /c/, as this is usually realized as a phonetic fricative.

Example II.14: contrasts between palatals

c vs. ɟ mácɛL debt gàcoj hoe

màːɟɛL dwarf gàꜜɟoj courageous man

c vs. ɲ coːn towards kocɛ bag

ɲon place koːɲɛŋ he curses

c vs. j càwɛŋ he sprinkles pàcɛŋ he carves

jàwɛŋ he shortens pàjɛŋ he vomits

12Note that I do not follow the Ethiopianist/Africanist tradition of transcribing the palatal

(22)

Consonants 71 ɟ vs. ɲ toɟɛŋ he disturbs ɟɔ̀nkɔ̀lɛ porcupine

toɲɛŋ he stops up ɲɔ̀ːnàL insult

ɟ vs. j toɟɛŋ he disturbs ɟàmɛ machete

toːjɛŋ he pierces jàwɛL circumcision

ɲ vs. j toɲɛŋ he stops up kàːɲ brideprice

toːjɛŋ he pierces kàːj night

II.2.2.4 Velars /k, g, ŋ/

The three velars /k, g, ŋ/ show minimal phonetic variation. Their pronuncia-tion corresponds to the IPA value of their respective symbols, except that the voiceless obstruent /k/ is pronounced with considerable aspiration: []. In the Godare variety of Majang, the voiceless plosive /k/ is often not fully re-leased at the end of a word, to the point that it is barely audible at all.

Example II.15: contrasts between velars

k vs. g korùŋ he shuts eːke ːr truth

gorùŋ he is sick eger how many?

k vs. ŋ kǎːrŋ he fights tàk inside

ŋǎːrŋ he goes tàŋL cow

g vs. ŋ dùgeŋ he hides tàgàL camel

dùːŋeŋ it evaporates tàːŋàːŋínL heron II.2.2.5 Nasals /m, n, ɲ, ŋ/

Because of their phonetic similarity, it is appropriate to show examples for contrasts between the four nasal phonemes of Majang.

Example II.16: contrasts between nasals

m vs. n cɛm straight dàːmɛ yellow

cɛːn pronoun 3S dàːnɛ beehive

m vs. ɲ màlɛŋ he strikes kɔ̀míL zebra

ɲàːlɛŋ it is light kɔ̀ɲɛŋ he persuades

m vs. ŋ mùkeŋ he stabs tɛm small

(23)

n vs. ɲ orpànL naming ceremony nomɛŋ he follows

pàːɲL mortar ɲon place

n vs. ŋ doːkùn bride-price nàcíL bread

dokùŋ it lands ŋàdíL pity

ɲ vs. ŋ ɲàːj liver koːɲ curse

ŋà ːj old woman ko*ŋ he gathers

Majang has an active nasal assimilation rule, which turns each alveolar nasal placed in front of a velar consonant into a velar nasal, as in àŋàn ‘four’ versus [àŋàŋ-k] ‘four (POSS)’.

II.2.3 Distribution

Before discussing the distribution of consonants, it is necessary to give some information on syllable, root and word structure in anticipation of sections II.4 and II.7. Syllables can be open and closed, and there can be two-conso-nant clusters both word-medially and word-finally. Two-consotwo-conso-nant clusters can also appear inside a morpheme, as in the verb stem dèrt ‘slide’. Words can be monomorphemic, but frequently include suffixes and even enclitics. Prefixes are very rarely encountered; these always end in a vowel. Noun and verb roots are often monosyllabic, but can consist of several syllables.

Most consonants can appear in all environments. There are, however, some limited distributions which are difficult to explain. Only a very small number of words end in a labial consonant, and except for the adverb otoːp ‘often’, and the noun ɟoːpL ‘people’, no words end in a labial obstruent. The labial nasal /m/ does appear word-finally in about 20 words, but this is balanced by the more than 200 words ending in the alveolar nasal /n/. Between vowels and in word-initial position /m/ is by far the most frequent nasal.

More transparent is the fact that at the end of a word, non-palatal voiced obstruents are not permitted phonetically. If an underlying voiced obstruent appears in that position, it is devoiced, such as íɟàːgL ‘work’, which is

(24)

Consonants 73 All consonants appear in good numbers word-initially and between vowels, except for the alveolar nasal /n/, which surprisingly appears word-initially in only about a dozen lexemes, as opposed to nearly ninety for /m/, 21 for /ɲ/ and 34 for /ŋ/. Except for the assimilation of alveolar nasals to the velar place of articulation (see section II.5.1), nasals keep their place of articula-tion, both inside the root and at morpheme boundaries: for example, ámɗL

‘stomach’ andɗàmɗú ‘after he ate’ show /m/ maintaining its labial place of articulation preceding alveolar phonemes. The two reduplicated verb stems tíntím ‘fold’ and tɔ̀ntɔ̀m ‘incubate’, however, suggest that some assimilation happens preceding alveolar consonants inside a stem. No other instances were encountered where any nasal assimilates to another place of articula-tion. tɛmk ‘it is small (SUB)’ shows /m/ stable in front of a velar consonant.

The alveolar nasal can be encountered preceding palatal consonants, as in pàrànɟíL ‘white person’. The palatal nasal was encountered preceding /k/ in

kóːɲkàn ‘cursed person’. The velar nasal was sometimes found preceding al-veolar phonemes, as in the stem cɔ̀ːlíláŋtL ‘vulture (NOM)’. It appears that

Majang can be characterized as a split system regarding nasal assimilation, with a clear assimilation of alveolar nasals towards velar nasals, and a gener-al lack of assimilation between other places of articulation.

In consonant clusters across syllable boundaries, there is a high tendency for having sonorants as first consonant (C1) and obstruents as second consonant (C2), as in words like tàmàːrɛrkí ‘they learn’ or, twice, in ɓànkàwkà ‘power’. Obstruents can only be found as C1 when C2 is also an obstruent, as in táːp ꜜ t áːnákL ‘letters’ or ìɟáːgkɛL ‘when they started to work’. In the same way,

sonorants also only appear as C2 when C1 is another sonorant, as in mérménáːꜜrá ‘I notice’ or ɓòkóːrjántáL ‘tortoise (DAT)’.

Implosives usually appear as C2 in a consonant cluster. An alveolar im-plosive /ɗ/ cannot be found preceding another consonant within the same word. It is changed to the flap /r/, as in eːŋàɗàːŋ ‘I smell’ vs. eːŋàrŋ ‘he smells’. The labial implosive /ɓ/ is rarely found preceding consonants. The only example encountered in the text corpus is gàgáɓkɛjgíɗáL whenever I

(25)

In the few cases of monomorphemic and polymorphemic word-final conso-nant clusters, C1 needs to be a [+SONORANT]13 consonant. If C1 is /l/ or a

nasal, then C2 can be a [-SONORANT] consonant (including implosives), or the

nasal /ŋ/ (such as in ɛltL‘grassland’, emɗ ‘canoe’ or tà*mŋ ‘it drips’). If C1 is a [-NASAL], [-LATERAL] sonorant /r, j, w/, then only /ŋ/ and /n/ can be found in

the C2 position, as in ko*wŋ ‘he digs’, kàːrnL ‘war’, ɓɛjnL ‘thorn’, or ɗɛjŋL

‘cooking stone’. The restrictions on word-final consonant clusters are entire-ly in line with the sonority hierarchy proposed by Lowenstamm (1981), which predicts the following order for syllable codas: approximant – sono-rant – nasal – obstruent. Two phonemes from the same level can combine to a consonant cluster. This is exactly what can be seen in Majang, giving the implosives a special status as being neither a sonorant nor an obstruent, placing them between nasals and obstruents.

The language strongly disfavors a syllable rhyme containing both a long vowel and a consonant cluster14. If these two things come together in any

configuration, the vowel is shortened. This can be observed for example in the locative form of ŋédàn ‘bee’, which makes use of the non-central case stem /ŋédàːn/, to which the locative stem extension -t is added, resulting in the form *ŋédàːnt, which is shortened toŋédànt.

II.3 Distinctive Features

When describing phonological rules, I occasionally make reference to the distinctive features of the phonological system. It is therefore necessary to determine what the distinctive features are for each phoneme. The discussion of the features presented here is not intended to be a contribution to any pho-nological model interested in features. No attempt is made to validate or dis-prove any existing feature inventories. The inventory proposed here is entire-ly language-specific and was chosen because it reflects the phonological realities of Majang. The vowels of Majang can be classified as seen in Table 4, with the understanding that they are all further marked as [+SYLLABIC]:

13See section II.3 for a definition of each distinctive feature.

14There are exceptions to this rule, such as the three possessive pronouns gáːnk (1

S.PL), goːnk

(26)

Distinctive Features 75 [-LABIAL] [+LABIAL] [+HIGH] i ù [-HIGH] [+ATR] e o [-ATR] ɛ ɔ̀[-BACK] [+BACK]

Table 4: distinctive features of Majang vowel phonemes

It may appear surprising that in such a feature representation the [-HIGH]

vowels are not further distinguished for vowel height, including the low vowel //. The introduction of the feature [LOW] would introduce an

unneces-sary redundancy into the feature system. But it needs to be pointed out that the representation in table 4 masks another lower-level redundancy: the fea-ture [-ATR] is neither required nor desired to characterize the vowel //. I prefer the use of the feature [LABIAL] over the feature [ROUND], because the

labial consonants and the round vowels trigger the same process, and there-fore it may be assumed that the same phonological feature is responsible. See section II.5.3 for a detailed discussion on how this feature [LABIAL]

gov-erns labial harmony.

The feature [ATR] may also be a surprising choice, as it was established that there is no ATR harmony in the language. Still the feature [ATR] is the only distinctive feature provided by the standard literature of generative phono-logy with the power to distinguish between the [-HIGH] vowels, even if a

fea-ture [LOW] were invoked, which would only apply to /à/ in Majang. Hall

(2007, p. 329) observes that “the feature [ATR] is used to capture the con-trast between /i e o/ ([+ATR]) and / / ([ɪ ɛ ɔ/ ([ ɛ ɔ/ ([ ɔ/ ([ -ATR]) – both in West African languages with [ATR] harmony, as well as Germanic languages like Eng-lish.” This is in agreement with Casali’s (2008, p. 507) explanation that [+ATR] vowels are characterized by a lower F1 value than [-ATR] vowels (and are therefore higher in traditional phonetic terms). The only other fea-ture previously used for this purpose, [TENSE], was abandoned in the early

days of generative phonology. Ladefoged’s (2005, p. 8 f) suggestion to use a feature [MID] for languages without ATR-based vowel harmony was not

(27)

restric-tions between some vowels; further traces can be seen in some morpheme al-ternations such as the nominalization markers -ɔ̀n and -àn (see section IV.2.2.1). If it was the feature [ATR] that caused this harmony in the past, it appears logical that the same feature still distinguishes the same vowels, even if a harmony is no longer based on this distinction.

[ATR] is the only vowel feature which is not also used for consonants. The following table helps to identify the consonant features, which are all needed in addition to the feature [-SYLLABIC].

[+LABIAL] [-LABIAL] [-HIGH] [+HIGH] p t c k [-VOICED] [+OBSTRUENT] b d ɟ g [+VOICED] [-OBSTRUENT] [-SONORANT] ɓ ɗ [+SONORANT] w r j [-LATERAL] [-NASAL] l [+LATERAL] [+NASAL] m n ɲ ŋ [-BACK] [+BACK]

Table 5: distinctive features of Majang consonant phonemes

The classification with respect to the manner of articulation follows Cle-ments and Osu (2002, p. 308), who state that “[...] the common property dis-tinguishing implosives from explosives is the absence of air pressure buildup in the oral cavity. [...] this property is exactly the correlate of the feature [-OBSTRUENT].” Indeed the two implosives have a different distribution from

both obstruents and sonorants in Majang, and therefore the [OBSTRUENT]

fea-ture provides the main dividing line between the Majang consonants. The eight [+OBSTRUENT] phonemes only need to be further classified as [-VOICED]

and [+VOICED]. For the ten [-OBSTRUENT] sounds, a further division is made by

the feature [SONORANT], which cannot just be seen as the inversion of the

fea-ture [OBSTRUENT] (Clements & Osu, 2002, p. 337f), but which in turn singles

out the implosives from the other [-OBSTRUENT] phonemes. The [+SONORANT]

phonemes are either nasals [+NASAL] or approximants [-NASAL]. The

(28)

Distinctive Features 77 the lateral approximant on the other display different phonotactic behavior (see section II.2.3).

The Majang consonant system provides no need for using the feature [CORONAL] to specify any place of articulation, as the feature [LABIAL] is needed

anyway in the language and therefore is sufficient to distinguish between labial and alveolar sounds. The labial harmony (section II.5.3) proves that the feature [LABIAL] is applied to both vowels and consonants; the labial

con-sonants and round vowels clearly form a natural class. This renders the fea-ture [CORONAL] redundant in Majang. The three remaining places of

articula-tion – alveolar, palatal, and velar – can easily be distinguished by the fea-tures [HIGH] and [BACK], which are also used for distinguishing vowels.

In total, all 25 consonants and vowels of Majang can be distinguished by the following ten non-prosodic distinctive features [SYLLABIC], [OBSTRUENT],

[SONORANT], [NASAL], [VOICED], [LATERAL], [LABIAL], [HIGH], [BACK] and [ATR].

To be able to fully accommodate all phonological phenomena of Majang in a feature notation, three prosodic features would have to be introduced: [LONG], a tone and a register feature. But apart from the feature [LONG],

proso-dic features are not used in this language description.

II.4 Syllable and Root Structure

II.4.1 Syllable structure

The Majang maximal-syllable template describing the possible phonetic syl-lables of the language is CVC, which allows for consonant clusters across syllable boundaries word-medially. Only at the end of a word this template can be exceeded by monosyllabic consonant clusters, but usually only fol-lowing short vowels15, and if they do not violate restrictions caused by the

sonority hierarchy. For more details on consonant clusters, see section II.2.3. Unlike Bender (1983, p. 115), the analysis of this study does not assume the presence of diphthongs in Majang. The examples given by Bender are

(29)

stead interpreted as sequences of a vowel and an approximant: kùroi (Bender) vs. kùꜜroj ‘donkey’; wàikùn (Bender) vs. wàjkùn ‘seed’.

In Majang there are many places where the transition from a syllable ending in a vowel to a syllable beginning with a vowel invites hearing a glide [j] or [w], depending on the nature of the vowels involved. This study follows the convention of only writing a glide between vowels if its underlying existence can be established from additional evidence, like if in a different morpholog-ical context the glide is found at the word boundary or in front of a conso-nant. Thus I write cáːkojɛ ‘valleys’ because of càːkoj ‘valley’, but ɓoɛ ‘antelope’, because in this word, the final /ɛ/ is part of the noun root, and there is no independent evidence for a glide between the two vowels, besides the possibility to transcribe this word phonetically as [ɓojɛ]. The final two syllables in the two words cáːkojɛ and ɓoɛ are phonetically not easily distin-guishable16.

In order to maintain the Majang syllable structure, the language uses a num-ber of devices to break up unacceptable consonant clusters. One of these, in-volving the presence of alveolar implosives preceding other consonants, is introduced as rule 3 in section II.5.2. Less specific is a vowel-epenthesis rule which inserts the default vowel /i/ between stems ending in a [-SONORANT]

consonant and a consonantal suffix or enclitic. A very rough representation of this rule might look like this:

Rule 1: vowel epenthesis

Ø i / C __ + C [-SONORANT]

This rule is applied for example with the subordinate-clause clitic =k and with the sentence-final topicality (SFT) clitic =ŋ, which attach themselves

without epenthesis to preceding vowels and most sonorants, but require this epenthetic vowel in a position following obstruents:

16With ɓoɛ a speaker may also choose to separate the two vowels by a glottal stop [ɓoʔɛ].

(30)

Syllable and Root Structure 79 Example II.17: contrasts between nasals

a) ɗùŋe ŋ < {ɗùŋeL=ŋ} b) kùꜜro jŋ < {kùꜜrojL=ŋ}

hyena\SG.NOM=SFT donkey\SG.NOM=SFT

c) cɔ̀ːlíláŋtí ŋ < {cɔ̀ːlíláŋtL=ŋ} d) kùrɓùtù ŋ < {kùrɓùt=ŋ}

vulture\SG.NOM=SFT maggot\SG.NOM=SFT In examples a) and b), there is no epenthetic vowel, because the SFT-clitic

finds a vowel or approximant (/j, w/) to attach itself to. Examples c) and d) have an epenthetic vowel following a stem ending in an obstruent. In ex-ample d), the epenthetic vowel changes from /i/ to // as a result of the labial-harmony rule 4, introduced in section II.5.3 below. This shows that the epenthetic vowel is subject to this rule which otherwise only applies to ver-bal suffixes.

The feature [-SONORANT] is a simplification of the actual environments

trig-gering this rule. Quite often the epenthesis also happens following nasals and even some other oral sonorants. For the SFT-clitic the following

generaliza-tions hold:

• A stem ending in a vowel or an approximant will never have the epenthetic vowel (as it would serve no purpose): ɗùmà ːŋ ‘owner (NOM)’, ɗùŋe ŋ ‘hyena (NOM)’, kùꜜro jŋ ‘donkey (NOM)’.

A stem ending in a [-SONORANT] sound will almost always have the

epenthetic vowel: weːntí ŋ ‘ear (NOM)’, émɗíŋ ‘canoe (NOM)’, ɗóːkù ŋ

‘land (NOM)’. Stems ending in an alveolar implosive which is not part

of a CC-sequence change this implosive to the phoneme /r/ (see sec-tion II.5.2). This explains the alternasec-tion between eːŋàɗàːŋL‘I smell’ and eːŋàrŋ ‘he smells’.

• Stems ending in a nasal use the epenthetic vowel only following a long stem vowel: dɔ̀mɔ̀ːnù ŋ ‘leopard (NOM)’, wàːjàːní ŋ ‘plant (NOM)’,

ɛdɛːní ŋ ‘mountain (NOM)’. Following a short vowel, the SFT-marker is

dropped entirely: tɔ̀ n ‘child (NOM)’, tèkán ‘aunt (NOM)’, càːkóm ‘friend

(NOM)’. This criterion looks more straightforward than it actually is.

One of these long vowels is apparently caused by the SFT-marker;

without it, the plain nominative form of ɛdɛn ‘mountain (ABS)’ is

ɛdɛnL with a short vowel. The

SFT-marker therefore seems to create

(31)

• For other sonorant environments, the placement of the epenthetic element appears to be even more random, as illustrated by word pairs such as kùtùr ‘hog (ABS)’ vs. kùtùrùŋ ‘hog (NOM)’ as opposed to

wàrL‘dog (

ABS)’ vs. wà rŋ ‘dog (NOM)’. For lack of a better

generaliza-tion, it needs to be concluded that stems ending in /l/ and /r/ can sometimes be found taking a vowel, and sometimes not.

II.4.2 Root structure

In Majang, the synchronic distinction between roots and stems is often difficult to make. Some derivation affixes clearly identify many stems as derived from particular roots, as in the case of ɓoːjànL ‘hate (INF)’, which is

derived from the verb root ɓoːj. There are, however, many lexemes where the different stem forms cannot be easily traced back to an identifiable under-lying root. This is the case when the differences between the stems are either of a purely tonal nature, or when there are differences in vowel length or quality, or when the stems differ in the absence or presence of a particular (usually stem-final) consonant, which cannot be identified as a productive derivation marker. All these factors can be illustrated by the stems ɛdɛn (SG.ABS), ɛdɛnL(SG.NOM.MOD), ɛdɛ ːn (SG.ERG), ɛdɛnL(SG.NOM), edeːn (SG.LOC), ɛdɛːn

(SG.DAT), ɛdɛnk (PL.ABS) and ɛdɛn (PL.ERG). These forms all refer to the lexeme

for ‘mountain’, and some of them have additional grammatical uses to those listed here. Each stem is treated as a simple stem in this study, which implies that they are not treated differently from a root. This is particularly the case when looking at the phonological structure of roots, as this section attempts.

II.4.2.1 Noun roots

(32)

Syllable and Root Structure 81 Example II.18: examples of noun roots

1-syllabic gloss 2-syllabic gloss 3-syllabic gloss

káːrn war kántɛ basket kácíkír stump

mòːr anvil móːrɛ fat mɛkɛlɛmL hawk

tèk in-law tálój swarm tɛŋɔ̀níː locust

wàː house (DAT) wòrí money wàːkójót god

ɓàːj heart bɔ̀dɛL palm tree bùcùlɛ puppy

II.4.2.2 Verb roots

Verb roots in Majang have a very strong tendency to end in a consonant. This is compatible with the fact that practically all person suffixes begin with a vowel. Roots ending in a consonant therefore enable easy syllabifica-tion. This is certainly true for all roots of the i- and -classes of verbs (seeɛ-class verb suffixes section IV.2.1.1 on verb classes). The various a-classes, however, contain many roots ending in a vowel. This may be the reason for the development of the -k extension morpheme used in these classes (see p. 248), which does not seem to carry any meaning. Its purpose appears to be to create well-formed stems which otherwise would end in a vowel.

Monomorphemic perfective verb stems usually consist of one syllable, but a smaller number of roots has two syllables.

Example II.19: examples of monomorphemic verb roots 1-syllabic gloss 2-syllabic gloss àːm yawn àgàl steal

càn lose ɓɔ̀lɔ̀ːr grow

dèrt slide dìgòj greet

gòt blow gùpàt spill

óːj shout ògàr cut hair

(33)

Example II.20: imperfective stem formation through reduplication

a) dɛnɛ dɛdɛn

{dɛn-ɛ} {dɛdɛn}

see\PFV-3S.CJ see\IPFV.3S.CJ

he sees he is seeing

b) ɗàm ɗàɗàmíːL {ɗàm} {ɗàɗàm-iːL}

eat\PFV.3S.CJ eat\IPFV-AP.3S

he eats he is eating

c) reːꜜràL rereràL {reːr-àL} {rerer-àL}

run\PFV-1S.DJ run\IPFV-1S.DJ

I run I am running

II.5 Phonological Processes

There are few phonological rules on segments which apply across the whole language. The two most regular rules are nasal assimilation and alveolar im-plosive weakening. Two other important rules are morphologically restricted to only some sets of suffixes: height assimilation and labial harmony.

II.5.1 Nasal assimilation

As seen in section II.2.2, alveolar nasals adapt their place of articulation when preceding velar consonants, as in àŋàn ‘four’ versus [àŋàŋk] ‘four’

(POSS). This was not observed for alveolar nasals preceding palatal or labial

consonants, for lack of words providing such an environment. Only alveolar nasals were found to assimilate their place of articulation. Section II.2.3 con-tains examples of possible clusters of nasals followed by non-homorganic consonants.

Rule 2: nasal assimilation n → [+BACK] / __ C

(34)

Phonological Processes 83 This rule states that /n/ changes to /ŋ/ preceding a velar consonant. It is diffi-cult to establish whether this rule is lexical, that is whether speakers are aware of this change (see Snider (2018, p. 95) for this component of the dis-tinction between lexical and postlexical rules). Throughout this study it is treated as postlexical, which means that it is not reflected by the ortho-graphic conventions adopted for this study. The possessive form of àŋàn is therefore written as àŋànk.

For many morphemes it is impossible to establish whether the phonetic com-bination [ŋk] goes back to an underlying cluster /nk/ or /ŋk/. This is true for many pronouns and determiners, such as gànk ‘POSS\1P.PL’, cenk ‘2S.CONTR’,

kɔ̀nk ‘REF\RECPST’, mànk ‘or’, or even the possessive case marker -onk. In all

these cases, I decided to transcribe them consistently as <nk>.

II.5.2 Alveolar implosive weakening

The alveolar implosive /ɗ/ is not allowed to occur preceding a morpheme beginning with a consonant. In that environment it is weakened to /r/, as in eːŋàɗàːŋL‘I smell’ vs.eːŋàrŋ ‘he smells’

.

A formal representation of this rule

might look as follows:

Rule 3: alveolar implosive weakening ɗ → r / __+C

There is no corresponding weakening rule for the labial implosive conso-nant.

II.5.3 Labial harmony

All verbal suffixes involving the short vowel /i/ in their first syllable are sub-ject to a vowel-harmony rule, which, depending on the rhyme (R)17 of the

preceding syllable, changes the suffix vowel. If the rhyme contains a [+LABIAL] consonant or vowel, the vowel is /ù/, otherwise it remains /i/.

(35)

Example II.21: labial harmony

a) díːꜜlínL b) làŋír c) pàːŋík {díːl-ínL} {làŋ-irL} {pàːŋ-ík}

take-2S.DJ find-3P.DJ slap-IMP.SG

you take they find slap!

d) ŋeɗeːmùnL e) ɓokotùr f) làptúk {ŋeɗeːm-ínL} {ɓokot-irL} {làpt-ík}

smile-2S.DJ kill-3P.DJ dive-imp.sg

you smile they kill dive!

If the last stem syllable contains a [+LABIAL] vowel or consonant anywhere in

the rhyme, the suffix vowel is //. This is the case in example e), where the labial vowel /o/ in the last stem syllable triggers the process. But, as stated, the process also applies to stems with a labial consonant in the coda, as seen in example II.21d). It is also important to note that the labial consonant trig-gering the harmony does not have to be the final consonant of the stem, but may be another consonant found in the coda, as the /p/ in example f). The nature of the onset of the preceding syllable has no impact on the choice of the suffix vowel, as can be seen from example c), where the onset consonant is [+LABIAL], but all rhyme sounds are [-LABIAL].

It is this labial harmony that prompted the choice of [LABIAL] as a feature

operating for both consonants and vowels in Majang (section II.3). As can be seen from examples d) and e), both a round vowel and a labial consonant de-termine that the suffix vowel changes from /i/ to //. It is best to assume that the rounding of the vowel is the result of the same phonological rule (not two different rules), and describing this one rule is easier when it makes re-ference to only one phonological feature. Another crucial assumption is that this feature [LABIAL] has a suprasegmental nature, determining that a whole

syllable rhyme takes over the feature [+LABIAL] when it is infected with it

anywhere. In the labial-harmony rule it is therefore necessary to make refer-ence to the syllable rhyme as a phonological unit:

Rule 4: labial harmony on verbal suffixes and epenthetic vowels

V → V / R + __

[+HIGH] [+LABIAL] [+LABIAL]

[-LABIAL]

(36)

Phonological Processes 85 This rule states that a suffix morpheme containing the short vowel /i/ changes it to // when following a morpheme ending in a syllable rhyme containing a labial sound. This does not apply to the long vowel //. The 1P

suffix -í,ːL in the verb ŋeɗeːmíːL ‘we smile’ is not affected by the rule. As stated in section II.1.3 above, it is not always easy to determine whether a particular suffix contains a long vowel or a short vowel. One of these suffixes is the class marker of i-class verbs in some complex verb forms, as in gáɓíːgíɗɛ {gáɓ-íː-gíɗ-ɛ} ‘it was given to’. If this were a short vowel, however, the labial-harmony rule would render this word as *gáɓúːgúɗɛ. The fact that this form does not materialize is taken as proof that this vowel of doubtful length must be a long one.

Still, it needs to be pointed out that this labial-harmony rule is restricted to verbal suffixes and epenthetic vowels. No nominal suffixes were encounter-ed that are subject to this phonological rule. The only exception to this is the possessive marker -k, which is only used on some nouns, and which is often introduced by the epenthetic vowel /i/. This epenthetic vowel is always sub-ject to labial harmony, also when it precedes the sentence-final topicality clitic =ŋ and the subordination clitic =k.

II.5.4 Vowel-height harmony for -class verb ɛ-class verb suffixes

Another phonological rule is restricted to verbs of the -class ɛ-class verb suffixes (see section -ɛ-class verb suffixes class verbs, p. 248 below); the class-specific verbal-suffix allomorphs that begin with a mid-front [-ATR] vowel /ɛ/ will change this vowel to its [+ATR] counterpart /e/ if the suffix follows a syllable containing a high vowel.

Example II.22: vowel height harmony on -class ɛ-class verb suffixes suffixes

a) 3S suffix {kɛːt-ɛ} he scatters {míːɲ-é} he covers

b) 2P suffix {kɛːt-ɛ*ːr} youPL scatter {míːɲ-e*ːr} youPL cover c) 3P suffix {kɛːt-ɛr} they scatter {míːɲ-er} they cover

d) INF, NEG {kɛːt-ɛːt} to scatter, scattering {míːɲ-eːt} to cover, covering This vowel-height harmony process is morphologically restricted to some suffixes applying on -class verbsɛ-class verb suffixes 18 and does not apply elsewhere in the

lan-guage.

18The impersonal marker -ɛꜜ, for example, is not affected by this rule, including when it

Referenties

GERELATEERDE DOCUMENTEN

(446) Comparison of the lowering of H triggered by surface Ls or floating Ms Once again these graphs highlight the fact that speakers AT and BL show patterns of phonetic

For this to happen, the surface tonal heights which are generated from phonological rules of consonant interaction are reanalyzed as underlying tones and those words coming from

Disambiguating Yoruba tones: At the interface between syntax, morphology, phonology and phonetics.. Tonal underspecification and Yoruba

ABS , PL -kùk plural plain absolutive marker for kùk L -plural nouns, and their plural formative for dative, locative and possessive case (IV.1.3.1) -n SG singular marker

After a short time with these men it became very clear that the Majang language had a surprise in store for me that I did not expect to find in an African language – a

It contains information about the location of the different language areas of Majang in relation to the areas of neighboring languages, the main roads in the area, the

The verb appears therefore in the (apparently un- marked) disjoint form, although the whole VP is conjoint (section IV.3.4.2). The conjoint-disjoint distinction is quite useful for

Not many generalizations about the case marking of unmarked singular nouns are possible. The ergative has low tones on short vowels and HL se- quences on long